Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.
com
To cite: Tamirou F,
Lauwerys BR, Dall’Era M,
et al. A proteinuria cut-off
level of 0.7 g /day after
12 months of treatment best
predicts long-term renal
outcome in lupus nephritis:
data from the MAINTAIN
Nephritis Trial. Lupus
Science & Medicine 2015;2:
e000123. doi:10.1136/lupus-
2015-000123
Received 4 September 2015
Revised 6 October 2015
Accepted 10 October 2015
For numbered affiliations see
end of article.
Correspondence to
Professor Frédéric A
Houssiau;
frederic.houssiau@uclouvain.
be
Tamirou F, Lauwerys BR, Dall’Era M, et al. Lupus Science & Medicine 2015;2:e000123. doi:10.1136/lupus-2015-000123
A proteinuria cut-off level of 0.7 g /day
after 12 months of treatment best
predicts long-term renal outcome
in lupus nephritis: data from the
MAINTAIN Nephritis Trial
Farah Tamirou,1 Bernard R Lauwerys,1 Maria Dall’Era,2 Meggan Mackay,3
Brad Rovin,4 Ricard Cervera,5 Frédéric A Houssiau,1 on behalf of the MAINTAIN
Nephritis Trial investigators
ABSTRACT
Background: Although an early decrease in
proteinuria has been correlated with good long-term
renal outcome in lupus nephritis (LN), studies aimed at
defining a cut-off proteinuria value are missing, except
a recent analysis performed on patients randomised in
the Euro-Lupus Nephritis Trial, demonstrating that a
target value of 0.8 g/day at month 12 optimised
sensitivity and specificity for the prediction of good
renal outcome. The objective of the current work is to
validate this target in another LN study, namely the
MAINTAIN Nephritis Trial (MNT).
Methods: Long-term (at least 7 years) renal function
data were available for 90 patients randomised in the
MNT. Receiver operating characteristic curves were
built to test the performance of proteinuria measured
within the 1st year as short-term predictor of long-term
renal outcome. We calculated the positive and negative
predictive values (PPV, NPV).
Results: After 12 months of treatment, achievement of
a proteinuria <0.7 g/day best predicted good renal
outcome, with a sensitivity and a specificity of 71%
and 75%, respectively. The PPV was high (94%) but
the NPV low (29%). Addition of the requirement of
urine red blood cells ≤5/hpf as response criteria at
month 12 reduced sensitivity from 71% to 41%.
Conclusions: In this cohort of mainly Caucasian
patients suffering from a first episode of LN in most
cases, achievement of a proteinuria <0.7 g/day at
month 12 best predicts good outcome at 7 years and
inclusion of haematuria in the set of criteria at month
12 undermines the sensitivity of early proteinuria
decrease for the prediction of good outcome. The
robustness of these conclusions stems from the very
similar results obtained in two distinct LN cohorts.
Trial registration number: NCT00204022.
INTRODUCTION
Despite significant therapeutic advances,
renal involvement remains a threatening
Brief communication
disease manifestation of systemic lupus ery-
thematosus, impacting quality of life and sur-
vival.1 Identification of short-term prognostic
factors predictive of poor long-term outcome
in lupus nephritis (LN) would be most
welcome, especially to fine-tune the intensity
and duration of immunosuppressive (IS)
therapy. Another compelling reason is to
determine optimal end points for clinical
trials so that new therapies can be evaluated
efficiently and accurately. In this respect,
several reports have demonstrated that an
early decrease of proteinuria after IS treat-
ment is predictive of a good long-term renal
outcome.2–5
Taking advantage of the long-term
follow-up data from the Euro-Lupus Nephritis
Trial (ELNT), we were able to determine the
optimal proteinuria target that maximises
sensitivity and specificity for prediction of
good long-term renal outcome. Thus, we
demonstrated that a proteinuria of <0.8 g/
day 12 months after randomisation was the
single best predictor of good long-term renal
function, with a sensitivity and specificity of
81% and 78%, respectively.6
The current analysis is aimed at testing the
validity of this proteinuria target in another
patient population with LN, taking advantage
of the long-term MAINTAIN Nephritis Trial
(MNT) data set.5 7
SUBJECTS AND METHODS
The MAINTAIN Nephritis Trial
The MNT is a European multicentre rando-
mised trial comparing azathioprine and
mycophenolate mofetil as maintenance IS
treatment of LN, after induction with
1
 Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.com

Lupus Science & Medicine

low-dose Euro-Lupus intravenous cyclophosphamide.

After long-term follow-up, renal relapse rates were
similar in the two arms, as published elsewhere.5 7

Patient selection
One hundred and five patients suffering from LN were
included in the MNT. For the purpose of this study, we
applied the same patient selection as for the corre-
sponding ELNT analysis,6 in order to make the two
cohorts comparable. Thus, 15 MNT patients were
excluded because serum creatine (sCr) measurement
was not available at or after 7 years of follow-up. Of note,
patients having achieved end-stage renal disease at any
time (n=4; at month 30, month 36, month 41 and
month 74) were included in the analysis. On the whole,
data from 90 MNT patients were studied.

Definition of good long-term renal outcome

Good long-term outcome was defined as sCr ≤1.0 mg/
dL at least 7 years after entry into the trial, again in
accordance with the criteria used for the ELNT analysis.6
Conversely, patients with sCr >1.0 mg/dL and those who
developed end-stage renal disease at any time were con-
sidered as having had a poor renal outcome.

Statistical analyses
We built receiver operating characteristics (ROC) curves,
through MedCalc, and calculated their area under the
curve (AUC; CIs) to test the performance of each pro-
teinuria level measured at month 3, month 6 and
month 12 as a predictor of long-term renal outcome.
Briefly, ROC curves plot sensitivity (true positive rate)
on the y axis against 1−specificity (false positive rate) on
the x axis. The point on the curve closest to the upper
left-hand corner, identified by the Youden index, corre-
sponds to the cut-off proteinuria value that optimises
sensitivity and specificity. The AUC summarises the
overall accuracy of a diagnostic parameter. AUC values
>0.9, >0.7 to 0.9, >0.5 to 0.7 and 0.5 are highly accurate,
moderately accurate, low accurate or equal to chance,
respectively.
The optimal proteinuria target value at month 12, as
defined supra, was used to calculate the positive and
negative predictive values (PPV and NPV). Figure 1 Receiver operating characteristic curves for
We also calculated the sensitivity, specificity, PPV and predictive value of 24 h proteinuria at month 3 (A), month 6
NPV with the addition of other clinical variables to pro- (B) and month 12 (C) of patients randomised in the
teinuria, including renal function and urinalysis. MAINTAIN Nephritis Trial. Sensitivity (true positive rate; y
axis) is plotted against 1−specificity (false positive rate; x
axis). The proteinuria (g/day) values indicated in the graphs
RESULTS optimise sensitivity and specificity. Figures in brackets are
Proteinuria levels that optimise sensitivity and specificity for 95% CIs. AUC, area under the curve.
prediction of good long-term renal outcome
We built ROC curves with proteinuria values measured
at different time points within the 1st year of treatment proteinuria cut-off values maximising sensitivity (CIs)
in order to identify the target that best predicts good and specificity (CIs). After 12 months of treatment,
long-term renal outcome. Figure 1 depicts the ROC achievement of a proteinuria <0.7 g/day predicted good
curves for proteinuria levels achieved at month 3, outcome, with a sensitivity and a specificity of 71% and
month 6 and month 12, their AUCs (CIs) and the 75%, respectively.

2 Tamirou F, Lauwerys BR, Dall’Era M, et al. Lupus Science & Medicine 2015;2:e000123. doi:10.1136/lupus-2015-000123
 Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.com

Brief communication

Table 1 Sensitivity, specificity, PPV and NPV for good long-term renal outcome according to target definition
Target at 12 months Sensitivity (%) Specificity (%) PPV (%) NPV (%)
Proteinuria <0.7 g/day 71 (48/68) 75 (9/12) 94 (48/51) 31 (9/29)
Proteinuria <0.7 g/day and sCr ≤1 mg/dL 63 (43/68) 83 (10/12) 96 (43/45) 29 (10/31)
Proteinuria <0.7 g/day and sCr ≤1 mg/dL and RBC ≤5/hpf 41 (28/68) 67 (8/12) 97 (28/29) 21 (8/38)
Figures in brackets are number of patients: for the sensitivity, patients with good outcome among those achieving the target/total patients with
good outcome; for the specificity, patients with poor outcome among those not achieving the target/total patients with poor outcome; for the
PPV, patients with good outcome/target achievers; for the NPV, patients with poor outcome/target non-achievers. Of note, only 80 of the 90
patients had proteinuria values available at month 12.
hpf, high power field; NPV, negative predictive value; PPV, positive predictive value; RBC, red blood cells; sCr, serum creatine.

Sensitivity, specificity, PPV and NPV
As indicated in table 1, the PPV for a good long-term
renal outcome of achieving a proteinuria <0.7 g/day at
12 months is very high (94%), thereby indicating that
proteinuria decrease alone drives long-term renal prog-
nosis. By contrast, the NPV is very low (31%), which
means that more than two-thirds of the patients not
meeting this target at month 12 will still experience a
good long-term renal outcome.
Next, we wondered whether adding renal function
and results of urinalyses would modify the sensitivity,
specificity, PPV and NPV. As shown in table 1, addition
of a normal renal function (sCr ≤1 mg/dL) to the
target criteria, did not significantly alter the results. By
contrast, requiring urine red blood cells (RBCs) to be
≤5/high power field (hpf ), in addition to proteinuria
<0.7 g/day and sCr ≤1 mg/dL as a response criteria
reduced the sensitivity from 71% to 41%, implying that
59% of the patients who will experience a good long-
term renal outcome would not be identified at month
12 if persistent haematuria is part of the target, while
only 29% would be missed if proteinuria alone is used
as criteria.
Last, we tested whether the NPV of proteinuria <0.7 g/
day at month 12 would be higher in patients presenting
with lower proteinuria levels at baseline. In these
patients, non-achievement of the target should be asso-
ciated with a poorer prognosis. We selected the mean
(3.3 g/day) and the median (2.4 g/day) proteinuria
values of the MNT cohort as cut-offs. As indicated in
table 2, the NPV is indeed higher in patients with a
lower baseline proteinuria level.
These observations are consistent with those derived
from the ELNT. Thus, Dall’Era et al demonstrated that a
cut-off proteinuria value <0.8 g/day at month 12 maxi-
mised sensitivity and specificity for the prediction of
good renal outcome and that addition of the results of
urinalysis negatively impacted early identification of
patients with good outcome, with a very comparable
drop in sensitivity from 81% to 47%.7 We suggest that
such a consistency across two distinct trials strengthens
the validity of our conclusions.
The finding that haematuria did not contribute as a
surrogate for long-term kidney outcomes is not so sur-
prising. In a multicentre trial, it is difficult to ensure
that the analysis of the urine sediment is done uni-
formly. The number of RBCs/hpf depends on the
volume in which the pellet of sediment is resuspended.
This is not standardised and if the volume is large the
cellular elements in the pellet will be diluted and the
counts inaccurate. The time of urine collection is
important, and how long urine sits before urinalysis may
affect the sediment, especially RBC casts. Additionally,
there are many reasons for haematuria in a population
of mainly young women. The red cells specific for glom-
erular bleeding, and thus indicative of LN, are dys-
morphic and are called acanthocytes. While these cells
are enumerated in centres that specialise in glomerular
diseases, in clinical trials all RBCs are counted, and
these are often eumorphic and from the lower urinary
tract. We do not discount the value of urine sediment
examination. If urine handling could be standardised
and the reader focused on the elements of the urine

DISCUSSION Table 2 NPV of a proteinuria <0.7 g/day at 12 months

Since permanent renal impairment is mostly a late event according to baseline proteinuria
in LN, only trials or cohorts with long-term follow-up Baseline proteinuria (g/day) NPV (%)
can unmask patients with poor renal outcome and allow
us to test the validity of short-term predictors, such as According to mean
<3.3 (n=48) 44 (7/16)
the kinetics of proteinuria decrease during the 1st year
≥3.3 (n=32) 15 (2/13)
of treatment. The data shown here, computed from the According to median
MNT, demonstrate (1) that achievement of a proteinuria <2.4 (n=37) 46 (6/13)
value <0.7 g/day at month 12 best predicts good renal ≥2.4 (n=43) 11 (3/16)
outcome at 7 years; and (2) that inclusion of microscopic Figures in brackets are numbers of patients, in casu poor
haematuria in the set of outcome criteria at month 12 outcomers/target non-achievers. Of note, only 80 of the 90
undermines the sensitivity of early proteinuria decrease patients had proteinuria values available at month 12.
NPV, negative predictive value.
for the detection of patients with good outcome.

Tamirou F, Lauwerys BR, Dall’Era M, et al. Lupus Science & Medicine 2015;2:e000123. doi:10.1136/lupus-2015-000123 3
 Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.com
Lupus Science & Medicine
that truly indicate glomerular injury, urinalysis may con-
tribute to a surrogate end point of renal response.
The first lesson learned from these analyses is that
assessment of microscopic haematuria should not be part
of the response criteria used in LN trials, in contrast to
the American College of Rheumatology (ACR) recom-
mendations,8 more so as interpretation of urinalyses is
complicated by delays in sample examination, the use of
automated techniques instead of microscopic reading
and interference by menstrual-related haematuria.
The second implication deals with treatment options
at the bedside after 1 year. Patients meeting the 0.7 g/
day proteinuria target at month 12 (ie, 64% of the MNT
cohort) can be reassured regarding their final renal
outcome, based on the very high PPV (94%). In such
patients, glucocorticoid tapering and withdrawal could
be considered, based on their well known contribution
to damage accrual,9 10 although only a controlled trial
could test the safety of this proposal in terms of renal
relapse rates. On the other hand, treatment decision will
be more difficult in patients not achieving the 0.7 g/day
target at month 12, due to the low NPV: two-thirds of
these patients will still experience a good outcome in
the long run. In this respect, the fact that the NPV is
twice as high in patients with a low baseline proteinuria
compared with those with higher values suggests that a
treatment switch might be more appropriate for the
former compared with the latter. The kinetics of protein-
uria decrease is obviously pivotal in this decision
process. Again, only a controlled trial could test the pos-
sibility that optimising therapy in partial responders
would be beneficial, a hypothesis tested in the RING
trial (RItuximab in lupus Nephritis with remission as a
Goal).
We acknowledge the limitations of our work. Data
were acquired in a small patient population. Most
subjects were Caucasians (79%) and suffered from their
first episode of LN (87%). Further work is needed to
improve the NPV which was not ameliorated by the use
of a more stringent target, taking into account
additional criteria, such as renal function or urinalysis.
Yet, the robustness of our conclusions stems from the
very similar results obtained in two prospectively
followed cohorts.
Author affiliations
1
Rheumatology Department, Cliniques Universitaires Saint-Luc, Pôle de
Pathologies Rhumatismales Inflammatoires et Systémiques, Université
catholique de Louvain, Bruxelles, Belgium
2
Division of Rheumatology, Russell/Engelman Research Center, University of
California, San Francisco, San Francisco, California, USA
3
The Feinstein Institute for Medical Research, Manhasset, New York, USA
4
Ohio State University, Wexner Medical Center, Columbus, Ohio, USA
5
Department of Autoimmune Diseases, Hospital Clinic, Universitat de
Barcelona, Barcelona, Catalonia, Spain
Collaborators This analysis was conducted as a collaborative work between
the MAINTAIN Nephritis Trial investigators and the Lupus Nephritis Trials
Network. The MAINTAIN investigators were: Daniel Abramowicz, Nephrology
Department, Hôpital Erasme, Université Libre de Bruxelles, Brussels,
Belgium; Fabiola Atzeni, Unita Operativa di Reumatologia, Ospedale Luigi
4 Tamirou F, Lauwerys BR, Dall’Era M, et al. Lupus Science & Medicine 2015;2:e000123. doi:10.1136/lupus-2015-000123
Sacco, Milan, Italy; Daniel Blockmans, General Internal Medicine Department,
UZ Gasthuisberg, Katholieke Universiteit Leuven, Leuven, Belgium; Maria
Giovanna Danieli, Istituto di Clinica Medica Generale, Universia Degli Study di
Ancona, Torrette di Ancona, Italy; Luc De Clercq, Rheumatology Department,
Sint-Augustinus Ziekenhuis, Wilrijk, Belgium; David D’Cruz, Louise Coote
Lupus Unit, St Thomas’ Hospital, London, UK; Maria del Mar Ayala
Guttierez, General Internal Medicine, Hospital Regional Universitario Carlos
Haya, Malaga, Spain; Enrique de Ramon Garrido, General Internal Medicine,
Hospital Regional Universitario Carlos Haya, Malaga, Spain; Inge-Magrethe
Gilboe, Rheumatology Department, Rikshospitalet University Hospital, Oslo,
Norway; Filip de Keyser, Rheumatology Department, UZ Gent, University of
Ghent, Ghent, Belgium ; Michel Delahousse, Service de Nephrologie, Hôpital
Foch, Paris, France; Gerard Espinosa, Department of Autoimmune Diseases,
Hospital Clinic, Barcelona, Catalonia, Spain; Christoph Fiehn, ACURA Center
for Rheumatic Diseases, Baden-Baden, Germany; Marc Golstein, Service de
Rheumatologie, Cliniques Saint-Jean, Brussels, Belgium; Loïc Guillevin,
General Internal Medicine Department, Hôpital Cochin, Paris, France; Marco
Hirsch, Luxembourg, Grand Duchy of Luxembourg; Alexandre Karras, Service
de Néphrologie, Hôpital Européen Georges Pompidou, Paris, France; Philippe
Lang, Nephrology Department, Hôpital Henri Mondor, Créteil, France;
Véronique le Guern, General Internal Medicine Department, Hôpital Cochin,
Paris, France; Martine Marchal, Service de Néphrologie, Hôpital de Tivoli,
La Louvière, Belgium; Antonio Marinho, Clinical Immunology Unit, Hospital
Santo Antonio, ICBAS, Porto, Portugal; Regina Max, Department of Internal
Medicine V, University of Heidelberg, Heidelberg, Germany; Patrick Peeters,
Nephrology Department, UZ Gent, University of Ghent, Ghent, Belgium; Peter
Petera, Zentrum für Diagnostik und Therapie rheumatischer Erkrankungen,
Krankenhaus Lainz, Wien, Austria; Radmila Petrovic, Institute of
Rheumatology, University of Belgrade, Belgrade, Serbia; Thomas Quémeneur,
Centre Hospitalier Régional Universitaire de Lille, Lille, France; Frank
Raeman, Rheumatology Department, Jan Palfijn Hospital, Merksem, Belgium;
Philippe Remy, Nephrology Department, Hôpital Henri Mondor, Créteil,
France; Isabelle Ravelingien, Rheumatology Department, Onze-Lieve-Vrouw
Ziekenhuis, Aalst, Belgium; Piercarlo Sarzi-Puttini, Unita Operativa di
Reumatologia, Ospedale Luigi Sacco, Milan, Italy; Shirish Sangle, Louise
Coote Lupus Unit, St Thomas’ Hospital, London, UK; Maria Tektonidou,
First Department of Internal Medicine, National University of Athens, Athens,
Greece; Lucia Valiente de Santis, General Internal Medicine, Hospital
Regional Universitario Carlos Haya, Malaga, Spain; Carlos Vasconcelos,
Clinical Immunology Unit, Hospital Santo Antonio, ICBAS, Porto, Portugal;
Luc Verresen, Nephrology Department, Ziekenhuis Oost-Limburg, Genk,
Belgium; Laurence Weiss, Département d’Immunologie, Hôpital Européen
Georges Pompidou, Paris, France; René Westhovens, Rheumatology
Department, UZ Gasthuisberg, Katholieke Universiteit Leuven, Leuven,
Belgium.
Contributors All coauthors have contributed to the study design, data
acquisition, data organisation, manuscript writing and reviewing.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Commission d’Ethique Hospitalo-Facultaire de l’Université
catholique de Louvain.
Provenance and peer review Not commissioned; externally peer reviewed.
Open Access This is an Open Access article distributed in accordance with
the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license,
which permits others to distribute, remix, adapt, build upon this work non-
commercially, and license their derivative works on different terms, provided
the original work is properly cited and the use is non-commercial. See: http://
creativecommons.org/licenses/by-nc/4.0/
REFERENCES
1. Costenbader KH, Desai A, Alarcón GS, et al. Trends in the
incidence, demographics, and outcomes of end-stage renal disease
due to lupus nephritis in the US from 1995 to 2006. Arthritis Rheum
2011;63:1681–8.
2. Houssiau FA, Vasconcelos C, D’Cruz D, et al. Early response to
immunosuppressive therapy predicts good renal outcome in lupus
 Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.com

nephritis: lessons from long-term followup of patients in the
Euro-Lupus Nephriti Trial. Arthritis Rheum 2004;50:3934–40.
3. Houssiau FA, Vasconcelos C, D’Cruz D, et al. The 10-year follow-up
data of the Euro-Lupus Nephritis Trial comparing low-dose and
high-dose intravenous cyclophosphamide. Ann Rheum Dis
2010;69:61–4.
4. Korbet SM, Lewis EJ, Collaborative Study Group. Severe lupus
nephritis: the predictive value of a ≥50% reduction in proteinuria at
6 months. Nephrol Dial Transplant 2013;28:2313–18.
5. Tamirou F, D’Cruz D, Sangle S, et al. Long-term follow-up of the
MAINTAIN Nephritis Trial, comparing azathioprine and
mycophenolate mofetil as maintenance therapy of lupus nephritis.
Ann Rheum Dis 2015;▪▪▪.
6. Dall’Era M, Cisternas MG, Smilek DE, et al. Predictors of long-term
renal outcome in lupus nephritis trials: lessons learned from the
Euro-Lupus Nephritis cohort. Arthritis Rheumatol 2015;67:1305–13.
Brief communication
7. Houssiau FA, D’Cruz D, Sangle S, et al. Azathioprine versus
mycophenolate mofetil for long-term immunosuppression in lupus
nephritis: result from the MAINTAIN Nephritis Trial. Ann Rheum Dis
2010;69:2083–9.
8. Renal Disease Subcommittee of the American College of
Rheumatology Ad Hoc Committee on Systemic Lupus
Erythematosus Response Criteria. The American College of
Rheumatology response criteria for proliferative and membranous
renal disease in systemic lupus erythematosus clinical trials. Arthritis
Rheum 2006;54:421–32.
9. Gladman DD, Urowitz MB, Rahman P, et al. Accrual of organ
damage over time in patients with systemic lupus erythematosus.
J Rheumatol 2003;30:1955–9.
10. Thamer M, Hernán MA, Zhang Y, et al. Prednisone, lupus
activity, and permanent organ damage. J Rheumatol 2009;36:
560–4.

Tamirou F, Lauwerys BR, Dall’Era M, et al. Lupus Science & Medicine 2015;2:e000123. doi:10.1136/lupus-2015-000123 5
 Downloaded from http://lupus.bmj.com/ on December 13, 2015 - Published by group.bmj.com

A proteinuria cut-off level of 0.7 g/day after 12

months of treatment best predicts long-term
renal outcome in lupus nephritis: data from
the MAINTAIN Nephritis Trial
Farah Tamirou, Bernard R Lauwerys, Maria Dall'Era, Meggan Mackay,
Brad Rovin, Ricard Cervera and Frédéric A Houssiau

Lupus Sci Med 2015 2:

doi: 10.1136/lupus-2015-000123

Updated information and services can be found at:

http://lupus.bmj.com/content/2/1/e000123

These include:

References This article cites 9 articles, 5 of which you can access for free at:
http://lupus.bmj.com/content/2/1/e000123#BIBL
Open Access This is an Open Access article distributed in accordance with the Creative
Commons Attribution Non Commercial (CC BY-NC 4.0) license, which
permits others to distribute, remix, adapt, build upon this work
non-commercially, and license their derivative works on different terms,
provided the original work is properly cited and the use is
non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Email alerting Receive free email alerts when new articles cite this article. Sign up in the
service box at the top right corner of the online article.

Notes

To request permissions go to:

http://group.bmj.com/group/rights-licensing/permissions

To order reprints go to:

http://journals.bmj.com/cgi/reprintform

To subscribe to BMJ go to:

http://group.bmj.com/subscribe/